Long-context Reasoning Accuracy on LongBench
70.4Accuracy (LongBench)CAP-CoT
Evaluation Results
| Method | Links | |
|---|---|---|
| CAP-CoTBackbone=GPT-4o2026.04 | 70.4 | |
| CAP-CoTBackbone=Deepseek-V32026.04 | 69.5 | |
| AoTBackbone=GPT-4o2026.04 | 69.5 | |
| CAP-CoTBackbone=GPT-4o-mini2026.04 | 69.3 | |
| CAP-CoTBackbone=Qwen-turbo2026.04 | 68.9 | |
| AoTBackbone=DeepSeek-V32026.03 | 68.7 | |
| AoTBackbone=Deepseek-V32026.04 | 68.7 | |
| AoTBackbone=GPT-4o-mini2026.03 | 68.5 | |
| AoTBackbone=GPT-4o-mini2026.04 | 68.5 | |
| AoTBackbone=Qwen-Turbo2026.03 | 68.1 | |
| AoTBackbone=Qwen-turbo2026.04 | 68.1 | |
| ECONBackbone=GPT-4o2026.04 | 66 | |
| ECONBackbone=Deepseek-V32026.04 | 65.2 | |
| GoTBackbone=GPT-4o2026.04 | 65.2 | |
| MADBackbone=GPT-4o2026.04 | 65.2 | |
| ECONBackbone=GPT-4o-mini2026.04 | 65 | |
| MADBackbone=Deepseek-V32026.04 | 64.5 | |
| ToTBackbone=GPT-4o2026.04 | 64.5 | |
| MADBackbone=GPT-4o-mini2026.04 | 64.2 | |
| ECONBackbone=Qwen-turbo2026.04 | 64 | |
| GoTBackbone=GPT-4o-mini2026.04 | 63.8 | |
| AFlowBackbone=GPT-4o2026.04 | 63.5 | |
| GoTBackbone=DeepSeek-V32026.03 | 63.4 | |
| GoTBackbone=Deepseek-V32026.04 | 63.4 | |
| ToTBackbone=DeepSeek-V32026.03 | 63.2 | |
| MADBackbone=Qwen-turbo2026.04 | 63.2 | |
| ToTBackbone=Deepseek-V32026.04 | 63.2 | |
| GoTBackbone=GPT-4o-mini2026.03 | 63.1 | |
| ToTBackbone=GPT-4o-mini2026.04 | 63.1 | |
| ToTBackbone=GPT-4o-mini2026.03 | 62.8 | |
| GoTBackbone=Qwen-Turbo2026.03 | 62.8 | |
| GoTBackbone=Qwen-turbo2026.04 | 62.8 | |
| ToTBackbone=Qwen-Turbo2026.03 | 62.4 | |
| ToTBackbone=Qwen-turbo2026.04 | 62.4 | |
| FoTBackbone=GPT-4o2026.04 | 61.5 | |
| AFlowBackbone=DeepSeek-V32026.03 | 61.4 | |
| AFlowBackbone=Deepseek-V32026.04 | 61.4 | |
| AFlowBackbone=GPT-4o-mini2026.03 | 61 | |
| AFlowBackbone=GPT-4o-mini2026.04 | 61 | |
| PromptAgentBackbone=GPT-4o2026.04 | 61 | |
| CoT-SCBackbone=GPT-4o2026.04 | 60.8 | |
| CCoTBackbone=GPT-4o2026.04 | 60.5 | |
| AFlowBackbone=Qwen-Turbo2026.03 | 60.4 | |
| AFlowBackbone=Qwen-turbo2026.04 | 60.4 | |
| Self-RefineBackbone=GPT-4o2026.04 | 60.2 | |
| TDA-RCBase LLM=DeepSeek-V32026.03 | 59.5 | |
| TDA-RCBackbone=DeepSeek-V32026.03 | 59.5 | |
| CoTBackbone=GPT-4o2026.04 | 59.5 | |
| TDA-RCBase LLM=GPT-4o-mini2026.03 | 59.3 | |
| TDA-RCBackbone=GPT-4o-mini2026.03 | 59.3 | |
| FoT (n=8)Backbone=DeepSeek-V32026.03 | 59.3 | |
| FoTBackbone=Deepseek-V32026.04 | 59.3 | |
| PromptAgentBackbone=Deepseek-V32026.04 | 59.2 | |
| FoT (n=8)Backbone=GPT-4o-mini2026.03 | 59.1 | |
| FoTBackbone=GPT-4o-mini2026.04 | 59.1 | |
| TDA-RCBase LLM=Qwen-Turbo2026.03 | 59 | |
| Instruction InductionBase LLM=DeepSeek-V32026.03 | 59 | |
| FoT (n=8)Backbone=Qwen-Turbo2026.03 | 59 | |
| TDA-RCBackbone=Qwen-Turbo2026.03 | 59 | |
| FoTBackbone=Qwen-turbo2026.04 | 59 | |
| CoT-SCBackbone=DeepSeek-V32026.03 | 58.9 | |
| CoT-SCBackbone=Deepseek-V32026.04 | 58.9 | |
| Instruction InductionBase LLM=GPT-4o-mini2026.03 | 58.8 | |
| CCoTBackbone=GPT-4o-mini2026.04 | 58.8 | |
| Role / Persona PromptingBase LLM=DeepSeek-V32026.03 | 58.7 | |
| PromptAgentBackbone=GPT-4o-mini2026.04 | 58.7 | |
| Role / Persona PromptingBase LLM=GPT-4o-mini2026.03 | 58.6 | |
| HoTBase LLM=DeepSeek-V32026.03 | 58.6 | |
| CoT-SC (n=5)Backbone=GPT-4o-mini2026.03 | 58.6 | |
| CoT-SCBackbone=GPT-4o-mini2026.04 | 58.6 | |
| Prompt CanvasBase LLM=GPT-4o-mini2026.03 | 58.5 | |
| Prompt CanvasBase LLM=DeepSeek-V32026.03 | 58.5 | |
| PromptAgentBackbone=Qwen-turbo2026.04 | 58.5 | |
| CCoTBackbone=Qwen-turbo2026.04 | 58.5 | |
| CCoTBackbone=Deepseek-V32026.04 | 58.5 | |
| HoTBase LLM=GPT-4o-mini2026.03 | 58.4 | |
| CoT-SC (n=5)Backbone=Qwen-Turbo2026.03 | 58.4 | |
| Self-RefineBackbone=DeepSeek-V32026.03 | 58.4 | |
| CoT-SCBackbone=Qwen-turbo2026.04 | 58.4 | |
| Self-RefineBackbone=Deepseek-V32026.04 | 58.4 | |
| Instruction InductionBase LLM=Qwen-Turbo2026.03 | 58.3 | |
| HoTBase LLM=Qwen-Turbo2026.03 | 58.2 | |
| Self-RefineBackbone=GPT-4o-mini2026.03 | 58.2 | |
| Self-RefineBackbone=GPT-4o-mini2026.04 | 58.2 | |
| Role / Persona PromptingBase LLM=Qwen-Turbo2026.03 | 58.1 | |
| Prompt CanvasBase LLM=Qwen-Turbo2026.03 | 58 | |
| Self-RefineBackbone=Qwen-Turbo2026.03 | 58 | |
| Self-RefineBackbone=Qwen-turbo2026.04 | 58 | |
| CoTBackbone=DeepSeek-V32026.03 | 57.7 | |
| CoTBackbone=Deepseek-V32026.04 | 57.7 | |
| CoTBackbone=GPT-4o-mini2026.03 | 57.6 | |
| CoTBackbone=GPT-4o-mini2026.04 | 57.6 | |
| CoTBackbone=Qwen-Turbo2026.03 | 57.3 | |
| CoTBackbone=Qwen-turbo2026.04 | 57.3 | |
| APBackbone=GPT-4o2026.04 | 55.2 | |
| Analogical PromptingBase LLM=DeepSeek-V32026.03 | 53.4 | |
| APBackbone=Deepseek-V32026.04 | 53.4 | |
| Analogical PromptingBase LLM=GPT-4o-mini2026.03 | 52.9 | |
| APBackbone=GPT-4o-mini2026.04 | 52.9 | |
| Analogical PromptingBase LLM=Qwen-Turbo2026.03 | 52.7 |